Is there a problem with the new Classical?


advanced search

Questions and Answers : Windows : Is there a problem with the new Classical?

Reply to this thread
Subscribe to this thread
Sort
AuthorMessage
adrianxw
Avatar
private message
Joined: Mar 8, 2006
Posts: 23
ID: 731
Credit: 245,048
RAC: 80
Message 452 - Posted 3 Apr 2006 10:25:19 UTC

End of this thread.
____________
Wave upon wave of demented avengers march cheerfully out of obscurity into the dream.

m.somers User profile image
Forum moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist
Avatar
private message
Joined: Nov 14, 2005
Posts: 662
ID: 1
Credit: 1,417,572
RAC: 2
Message 453 - Posted 3 Apr 2006 11:48:48 UTC

No not that I\'m aware of. I think he has another problem (hr_class related http://boinc.gorlaeus.net/forum_thread.php?id=71#413)...

m.
____________
M.F. Somers

adrianxw
Avatar
private message
Joined: Mar 8, 2006
Posts: 23
ID: 731
Credit: 245,048
RAC: 80
Message 454 - Posted 4 Apr 2006 8:41:53 UTC

That \"he\" is me. Since the change of \"Classical\" on 29th I have not had a single wu classed as \"Valid\". They run normally here, upload, and then get flagged as invalid. I notice others using top end gear are not having troubles, but their machines are being paired with similar equipment. My 2.533GHz machine is being paired with 4-500MHz machines.

I am well aware that a substantial difference in CPU can affect results, but I have no control over what gets the same wu\'s as me. The fact that the single non \"Classical\" unit I received worked tends to suggest an issue with Classical.

I have stopped crunching them now, it is simply wasting cycles.
____________
Wave upon wave of demented avengers march cheerfully out of obscurity into the dream.

m.somers User profile image
Forum moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist
Avatar
private message
Joined: Nov 14, 2005
Posts: 662
ID: 1
Credit: 1,417,572
RAC: 2
Message 455 - Posted 4 Apr 2006 12:34:40 UTC

Fact:

Your CPU type is specified as \"GenuineIntel x86 Family 15 Model 2 Stepping 4 2533MHz\". Apparently with your Windowze NT, the CPU naming is different than people using i.e. XP. They get CPU designations like \"Pentium 4\" or \"Pentium D\" or whatever...

Problem:

The current code looks at the CPU designation, a.k.a. the string it receives from your system, and then tries to classify this into a hr_class. If your system uses a different naming convention, this mapping obviously goes wrong, if you do not take this into account.

Fix ?!:

The obvious fix is to know how each (Windowze) system names each CPU and the make the hr_class mapping algorithm take care of all these possibilities... (I can obviously only do this as soon as I know which Windowze calls what CPU by what name, in the past and in the future.)

Real Problem:

However, clearly, this is not a true fix cause MicroLoft turns out to be inconsistent in naming the CPU\'s whereas, in my humble opinion, the name+brand of a CPU should be system independent (as it is the case in all Linux versions and flavors)!

So:

In other words, I could try and figure out how Windowze NT names all the CPU\'s and, how Windowze 98 does, and how Windowze 95 and and ... and then add all those names to the algorithm, knowing in advance that I have to do that again for each new CPU comming out or as soon as MicroBrain decides to change stuff again...

and...

all that for the few people running Windowze NT (20 of the 2089 windows hosts in my database) or Windows 98 (60 of the 2089) or whatever (in total about 3% off all the windows hosts)... as you can imagine not really on the top of my priority list ;-).

Moreover, at the moment less than 2% of the work gets invalidated incorrectly because of stuff like this. That\'s a *very* nice figure, if you understand how complicated things are...

m.
____________
M.F. Somers

adrianxw
Avatar
private message
Joined: Mar 8, 2006
Posts: 23
ID: 731
Credit: 245,048
RAC: 80
Message 456 - Posted 4 Apr 2006 14:54:00 UTC

I understand your problem.

Also, understand mine, I support software that is running on NT4 machines, therefore I need to have an NT4 box to ensure software modifications work on my clients machines. That NT4 system is not going away any time soon. Similarly, I have no control over how MS report the processor type.

Since the other BOINC projects, (Predictor, Rosetta, MalariaControl, SIMAP and Einstein), attached on that box run fine, I\'ll just drop out of Leiden. If I can see a way, maybe I\'ll try to find a few percent on one of my XP servers.

Best wishes for the future of your project.
____________
Wave upon wave of demented avengers march cheerfully out of obscurity into the dream.

Reply to this thread

Questions and Answers : Windows : Is there a problem with the new Classical?



Return to Leiden Classical main page


Copyright © 2017 Leiden University - Leiden Institute of Chemistry - Theoretical Chemistry Department